An Automatic PCB Imposition Method based on Reinforcement Learning

The growing complexity of electronic products demands more intricate Print Circuit Board (PCB) shapes. The complex-shaped PCBs often contribute to reducing efficiency in imposition process and result in material wastage. Currently, PCB imposition heavily relies on the personal experience of engineers, which is inherently uncertain and challenging to optimize. Additionally, the presence of specific PCB imposition rules leads to traditional optimization methods unsuitable for a direct solution. To address these issues, we propose an automatic PCB imposition method based on reinforcement learning. Our approach transforms the PCB imposition problem into a Markov Decision Process(MDP), decomposing the combination optimization problem into a sequential solution problem for the optimal placement of each PCB. We derive the global optimal solution by solving the optimal placement of each PCB. According to the simulation results, our approach demonstrates a 28.40% increase in utilization rate when compared to the random imposition algorithm and a 7.75% increase when compared to the greedy search algorithm.